One should assess whether the granularity of each dimension is actually required for the planning process. As a matter of fact, due to their nature, G/L actuals may often hold a huge amount of details which is not required for planning.
Workaround: An alternative design could also be to load detailed actuals for analytics purposes (if at all necessary) in a dedicated model, and then leverage a more condensed planning model for Budgets, Plans and Forecast. Alternatively, transaction details my continue to reside in the source system, and a drill-through process can be used from within the SAC story to see the additional details when required.
In additional to dimension granularity, which is typically concerned with the number of dimension members and depth of a hierarchies within dimensions, the number of dimensions within a model is also important. Excessive dimensionality makes overall model navigation more challenging, can create additional complexities for data entry (e.g. data validation), and often results an unnecessarily large and sparse data models. It is advisable then to try and keep planning models to between 8-12 dimensions (inclusive of standard dimensions such as account, date, and version). Where models exceed this recommendation, use of alternate hierarchies, navigational attributes, and commenting (via column comments) should be evaluated by the solution designer to rationalize the dimension count.
When starting a data input session for the first time, public data is – behind the scenes – cached to a dedicated private version for that user. This may result in huge amount of data being copied.
Workaround: a private version can be created beforehand, with the appropriate filters, so that only a subset of data is copied but not all. Working with such a private version will result in very noticeable performance gains when entering new data.
As a general recommendation, one must insist that every new private version must be created with filters. Then, the more discriminating these filters, the better.
On the fly calculations provided by account formula is an extremely powerful feature. However, this requires processor intensive cube calculations, especially when formulas are either numerous, or with nested calculations (one formula depends on the result of another formula, and so on). As a result, performance may be impacted by such circumstances.
Workaround: You can consider using Advanced Formulas, within data action, which are not executed in real-time, but only at user request. This may prove to reduce the overall system burden (see also the calculations guide in this series).
A natural behavior of a multi-dimensional engine is to sum up data along multiple hierarchies. If you think about it, the number of aggregations performed by the engine in any model is quite phenomenal. Suppose 2 dimensions with hierarchies that are 9 levels deep (with 10 leaf members). This already creates 100 potential aggregations. Fortunately, the SAP HANA database that SAP Analytics Cloud is leveraging can perform these calculations at the speed of light. However, when it comes to specific aggregations, such as averages or min/max values, etc.. these calculations require not only that standard aggregations are performed, but also specific aggregations get calculated in a second pass. This will result in a performance hit.
Workaround: some of these calculations may be moved to advanced formulas.
Zeros can actually be stored and are different from null values. Once in a while, you may delete zeros from the planning database by using native system functions in the modeler to delete facts. NB: in case you create advanced formulas for which you have to reset to zero some values prior to the new calculation to take place, please consider assigning a NULL value to the data region you wish to erase, and not assigning zeros to it (please see calculations guide)
Data access profiles do have an impact on story/query performance. As a general rule of thumb, the more restrictive a user’s data access, the better the performance. The rationale behind this is that a query output will only return data that a user is allowed to view or edit. However, the complexity and size of data access configuration may have an impact for the functional administrator as modeling rights will be refined depending on a power user privileges.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
30 | |
23 | |
10 | |
7 | |
6 | |
6 | |
5 | |
5 | |
5 | |
4 |